Skip to content

Proper-Time, Zeta, and Determinant Prescriptions

Proper-time, zeta-function, and spectral-product formulas describe the same one-loop object only after they use the same operator, domain, omitted modes, scale, analytic continuation, and local subtraction. Their formal symbols are short; the equivalence conditions carry the physics.

Required background. One-Loop Matter Effective Actions in Curved Space fixes the matter Hessian, and Heat Kernels, Zeta Functions, and Spectral Determinants supplies Mellin transforms and meromorphic continuation.

Helpful background. Spectra, Resolvents, Spectral Measures, and Functional Calculus clarifies domains and spectral cuts; Dimensional Regularization and Minimal Subtraction provides a comparison scheme.

Three definitions and one spectral problem

Section titled “Three definitions and one spectral problem”

Let L\mathcal L be a positive self-adjoint Laplace-type operator on a compact Riemannian manifold, with discrete eigenvalues λj>0\lambda_j>0 after removing an n0n_0-dimensional kernel. Then

K(s)=TresL=λj>0esλj,K'(s)=\operatorname{Tr}'e^{-s\mathcal L} =\sum_{\lambda_j>0}e^{-s\lambda_j},

and, initially for Rez\operatorname{Re}z large,

ζL(z;μ)=λj>0(λjμ2)z=μ2zΓ(z)0dssz1K(s).\zeta_{\mathcal L}(z;\mu) =\sum_{\lambda_j>0}\left(\frac{\lambda_j}{\mu^2}\right)^{-z} =\frac{\mu^{2z}}{\Gamma(z)} \int_0^\infty\mathrm ds\,s^{z-1}K'(s).

Meromorphic continuation to z=0z=0 defines

lndetLμ2=ζL(0;μ).\ln\det{}'\frac{\mathcal L}{\mu^2} =-\zeta'_{\mathcal L}(0;\mu).

The regulated proper-time expression removes the same short-ss asymptotic terms before taking its cutoff away. Hawking’s construction makes the zeta derivative and heat-kernel Mellin transform explicit for curved-space Gaussian integrals Hawking 1977, pp. 133–140; a modern convention-level statement appears in Vassilevich 2003, Eqs. (2.23)–(2.34).

Zero modes do not disappear mathematically. The prime means that the determinant is taken on (kerL)(\ker\mathcal L)^\perp; collective coordinates, gauge volume, or a constrained source integral supplies the missing finite-dimensional factor. Introducing a small mass and then taking it to zero gives

lndet(L+m02)=n0lnm02+lndetL+O(m02),\ln\det(\mathcal L+m_0^2) =n_0\ln m_0^2+\ln\det{}'\mathcal L+O(m_0^2),

so failing to subtract n0lnm02n_0\ln m_0^2 is an observable infrared error, not a choice of ultraviolet scheme.

First application: a compact ultrastatic spectrum

Section titled “First application: a compact ultrastatic spectrum”

Take the spatial operator on a circle of circumference LL,

L=d2dx2+m2,λn=(2πnL)2+m2,nZ.\mathcal L=-\frac{\mathrm d^2}{\mathrm dx^2}+m^2, \qquad \lambda_n=\left(\frac{2\pi n}{L}\right)^2+m^2, \quad n\in\mathbb Z.

For m>0m>0 it is positive. Poisson resummation gives

K(s)=L4πsem2sqZeq2L2/(4s).K(s)=\frac{L}{\sqrt{4\pi s}}e^{-m^2s} \sum_{q\in\mathbb Z}e^{-q^2L^2/(4s)}.

The q=0q=0 term is the local infinite-volume contribution; q0q\ne0 terms are exponentially small as s0s\downarrow0 and contain the global finite-size information. Subtract the same q=0q=0 term in proper time that analytic continuation removes in the zeta prescription. Differentiating the continued zeta function or integrating the subtracted heat kernel then yields the same finite nonlocal part,

lndetLdetL=2ln(1emL)\ln\frac{\det\mathcal L}{\det\mathcal L_{\infty}} =2\ln\bigl(1-e^{-mL}\bigr)

up to the explicitly chosen local normalization. In the massless case, n=0n=0 becomes a zero mode and the primed determinant must be used before the limit.

This example exhibits the division cleanly: short proper time fixes local subtraction, whereas winding sectors fix global finite-size dependence. A finite list of Seeley–DeWitt coefficients cannot reconstruct emLe^{-mL}.

Negative eigenvalues and determinant phase

Section titled “Negative eigenvalues and determinant phase”

Suppose one eigenvalue crosses from +λ+|\lambda_-| to λ-|\lambda_-|. A logarithm with cut angle θ\theta gives

lnθ(λ)=lnλ+iArgθ(1).\ln_\theta(-|\lambda_-|) =\ln|\lambda_-|+i\,\operatorname{Arg}_\theta(-1).

Changing the cut changes the phase by an integer multiple of 2πi2\pi i but does not change ordinary renormalization-scale dependence. The phase can encode a physical in–out instability only when the contour is fixed by the vacuum amplitude and agrees with an independent mode or tunneling calculation. A Euclidean negative mode by itself may instead diagnose that the chosen saddle is not a minimum.

For products of noncommuting pseudodifferential operators, zeta determinants need not satisfy detζ(AB)=detζAdetζB\det_\zeta(AB)=\det_\zeta A\det_\zeta B. Any multiplicative anomaly is a prescription-dependent local term controlled by the symbols; factorizing an operator therefore requires a check rather than an algebraic assumption.

The structure map locates proper time and zeta as parallel evaluations of one declared spectrum. The reader should trace both paths through the same zero-mode and subtraction boxes.

Proper-time integration and zeta differentiation agree only after sharing the spectrum, omitted modes, scale, and subtraction

Heat traces and zeta functions are Mellin-related representations of one spectral problem; global finite terms survive beyond the local short-time coefficients. Schematic; not to scale.

The compact positive case licenses a real primed determinant. Noncompact volume divergences require a relative determinant or density; continuous spectra require a spectral measure; boundaries require a self-adjoint elliptic domain; zero modes require collective-coordinate data; negative modes require a cut. Compare these cases in Domain and failure conditions.

The failure map separates three errors often merged into one: retaining a zero eigenvalue makes the determinant vanish, changing a cut changes a phase, and changing μ\mu changes local finite terms. They need different repairs.

A zero mode, negative mode, and subtraction-scale change produce distinct determinant failures and cannot be repaired interchangeably

Kernel removal, spectral-cut choice, and ultraviolet renormalization are independent parts of a determinant prescription. Schematic; not to scale.

Show from the small-m0m_0 eigenvalue product that det(L+m02)=m02n0detL[1+O(m02)]\det(\mathcal L+m_0^2)=m_0^{2n_0}\det{}'\mathcal L\,[1+O(m_0^2)].

Solution

Split the product into the n0n_0 zero eigenvalues and the positive spectrum. Each zero eigenvalue contributes m02m_0^2. For every positive λj\lambda_j, factor λj+m02=λj(1+m02/λj)\lambda_j+m_0^2=\lambda_j(1+m_0^2/\lambda_j). After the same ultraviolet regularization used for the primed product, the second factor tends to one, giving the stated result.

Heat Kernels and the Schwinger–DeWitt Expansion develops the short-time coefficients. Imaginary Effective Actions and Vacuum Instability adds the in–out interpretation of a phase.

  • Hawking, Stephen W. “Zeta Function Regularization of Path Integrals in Curved Spacetime.” Communications in Mathematical Physics 55 (1977): 133–148. DOI.
  • Vassilevich, Dmitri V. “Heat Kernel Expansion: User’s Manual.” Physics Reports 388 (2003): 279–360. DOI. Open PDF.